Breast Carcinoma–Associated Fibroblasts Rarely Contain p53 Mutations or Chromosomal Aberrations Free

Epithelial cancers are markedly heterogeneous, composed of carcinomatous epithelial cells and varying proportions of a stroma that contains fibroblasts (the predominant cell type), endothelial, adipose, neural, and inflammatory cells. Although most of the cancer research done thus far has focused on the study of the carcinomatous epithelium, attention has recently centered on the essential interplay of the tumor with its microenvironment in regulating the biology of cancer cells. Differences in tumor stroma compared with normal stroma have been widely observed by pathologists. Bissell's group and others (1, 2) have reported that the phenotype of breast carcinoma–associated fibroblasts (CAF) is different from fibroblasts associated with normal mammary epithelium. Several studies have shown that CAFs exhibit cancer-promoting properties through a variety of different mechanisms (3, 4). Little was known about the underlying basis for such phenotypic differences, spurring investigations into the nature of genetic and genomic changes in CAFs as a possible rationale for this tumor-promoting phenotype. Experimental mouse models have shown that the presence of DNA changes in CAFs could lead to the acquisition of cancer-promoting capabilities. Hill and colleagues (5) reported that stromal cells carrying TP53 mutations are selected for by interaction with the cancerous epithelium in solid tumors, thereby facilitating stromal cell hyperproliferation. Furthermore, various human cancer cell lines have been shown to possess the ability to induce dozens of genomic changes—deletions and amplifications—in xenograft-associated host mouse stroma (6). Several reports demonstrating that loss of heterozygosity (LOH) is a common occurrence in cancer-associated stroma (ref. 7, and reviewed in ref. 8) in primary human breast cancers also seem to support the concept of genomic instability in tumor stromal cells. For instance, Patocs and colleagues (9) have shown the presence of somatic TP53 mutations and LOH in the cancer-associated stroma in one quarter of invasive ductal breast carcinomas, whereas no TP53 abnormalities were found in the corresponding cancerous epithelium. In contrast, however, a recent report by Qiu and colleagues (10) has challenged the notion of stromal coevolution. They profiled copy number and LOH in cultured and frozen breast and ovarian carcinoma–associated stromal samples and found only very rare copy number alterations and no TP53 mutations. The reason for this apparent contradiction is unclear, although it has been suggested that the former findings could have possibly arisen as a result of DNA artifacts that occurred as a consequence of paraffin fixation of tissue samples (11). In light of the lack of consensus pertaining to the nature of DNA copy number changes in the cancer-associated stroma, we have carried out comprehensive DNA copy number profiling and TP53 mutation analyses in cultured CAFs. In the largest series reported to date (n = 25), we found that both DNA copy number alterations and TP53 mutations are a very infrequent phenomenon (<5%) in human breast CAFs.

Invasive breast carcinoma specimens were surgically resected from patients at the Sir Mortimer B. Davis Jewish General Hospital (Montreal, Canada). Informed consent was obtained to bank breast tumor samples for research purposes. The clinicopathologic features of the breast cancers from which the CAFs were grown are shown in Supplementary Table S1. The collection of breast tumor surgical samples for research purposes was approved by the Ethics Committee of the Jewish General Hospital. Specimens were determined to be intratumoral by a breast pathologist and subsequently minced with a sterile blade and resuspended in a solution of DMEM with 10% fetal bovine serum (FBS) and 3% collagenase overnight at 37°C. The next day, samples were filtered through an 8-μm mesh to remove undigested debris. The single cell suspension with viable fibroblasts was cultured in DMEM (10% FBS) for 2 to 3 weeks in a 24-well plate and then transferred to a T75 flask where it was continually maintained. All fibroblasts were harvested between passage doublings five to eight. The counterpart fibroblasts from the T38 patient were taken at a distance of >2 cm from the invasive edge of the tumor. The MCF-7 breast cancer cell line was purchased from American Type Culture Collection and maintained in DMEM containing 10% FBS.

Cells were grown to 70% confluency and fixed for 20 minutes with formalin. Cells were then permeabilized with 0.2% Triton X-100 for 15 minutes followed by treatment with 3% hydrogen peroxide for 15 minutes to block endogenous peroxidases. Antibodies were subsequently allowed to incubate over a 1-hour period at room temperature; mouse anti-vimentin V9 clone (DakoCytomation), pan-cytokeratin antibody (NeoMarkers) specific for cytokeratins 1, 2, 5, 6, 7, 8, 11, 14, 16, 17, and 18, or mouse IgG₁ (R&D Systems). An anti-mouse secondary antibody (Cell Signaling Technologies) was then added at a 1:500 dilution followed by streptavidin-horseradish peroxidase (Cell Signaling Technology) at a 1:20,000 dilution. After washing, cells were treated for 15 minutes with a 3,3′-diaminobenzidine peroxidase substrate solution (Vector Laboratories) prepared according to the instructions of the manufacturer. The 3,3′-diaminobenzidine solution was washed thoroughly and cells were treated with Permount (Fisher Scientific) and stored at 4°C.

This protocol was carried out as recommended by the manufacturer (Agilent Technologies). All reagents were obtained from the manufacturer unless stated otherwise. Briefly, 5 μg of genomic DNA was digested overnight at 37°C with 1 unit/mL of both AluI and RsaI restriction endonucleases (Stratagene). The reaction was terminated by incubating at 65°C for 20 minutes. Digested DNA was subsequently incubated for 2 hours with the fluorescence Cy5 cyanine dye (in the case of fibroblast DNA) or fluorescence Cy3 cyanine dye (in the case of the female reference DNA, purchased from Stratagene) in the presence of deoxynucleotide triphosphates and exo-klenow polymerase. This reaction was terminated by a 20-minute incubation at 65°C. Labeled DNA was cleaned up using YM-30 Microcon columns (Millipore) according to the directions of the manufacturer. Labeled Cy3 and Cy5 DNA were combined and diluted in array comparative genomic hybridization (aCGH) buffer and Cot-1 DNA (Invitrogen). Arrays were allowed to hybridize for 40 hours at 65°C before being washed in a solution of 0.5× SSPE, 0.005% N-lauroylsarcosine, followed by 0.1× SSPE, 0.005% N-lauroylsarcosine, and scanned (Agilent DNA Microarray Scanner) at a resolution of 5 μm. All data were extracted from raw images and normalized using the Feature Extraction software version 9.5 (Agilent Technologies).

Genomic copy number differences were inferred using circular binary segmentation (CBS) analysis that searches for particular change points at which neighboring regions of DNA exhibit a statistical difference (P < 0.05) in copy number (12). In this analysis, each chromosome was divided into contiguous regions of equal copy number with a SD cutoff of 1.00. CBS was implemented with the DNA copy number package for the R statistical computing environment (http://www.bioconductor.org/). The graphic visualization of aCGH segmented data was also implemented in the R language.

The human CAFs, T38 CAF and CHUM496, were cultured in DMEM supplemented with 20% FCS. Cells were incubated in 5% CO₂. Harvesting, metaphase preparation, and cytogenetic analysis with a trypsin-Giemsa banding technique were performed according to standard cytogenetic procedures. Clonal chromosomal abnormalities and GTG-banded karyotypes were described according to the International System for Human Cytogenetic Nomenclature (13).

Two methods were used to screen for mutations in TP53. Mutations were initially screened by single-strand conformation polymorphism analysis of exons 5 to 9 of DNA samples as previously described (14). Then, DNA sequencing of exons 2 to 11, which corresponds to all protein-encoding regions of the gene, was then performed to identify mutations missed by this initial screen, as previously described (15). The primer sets and conditions for DNA sequencing are shown in Supplementary Table S2. Sequencing reactions were performed at the McGill University and Genome Quebec Innovation Centre (Montreal, Quebec, Canada). The TP53 mutation sequence variants were classified based on information from the IARC TP53 Database (16), relative to nucleotide position based on the GenBank reference sequence NC_000017 (17). LOH analysis of the TP53 locus was performed using the polymorphic microsatellite repeat marker D17S786, which maps ∼1.2 Mb proximal to TP53, as previously described (14).

Expression profiling was carried out according to the instructions of the manufacturer (Agilent Technologies). Briefly, fibroblasts were harvested from subconfluent cultures, cultivated in DMEM with 2% FBS. RNA was then extracted using the Mini RNA Extraction kit (Qiagen Sciences). Five micrograms of total RNA was reversed transcribed with the Fairplay III Microarray Labeling kit according to the instructions of the manufacturer (Stratagene). The resulting cDNA was then precipitated with 70% ethanol, air-dried, resuspended in 5 μL of coupling buffer, and dissolved at 37°C for 15 minutes. Five microliters of Cy3 or Cy5 dye were added to the universal reference (Stratagene) or fibroblast cDNA, respectively, and allowed to incorporate for 30 minutes at room temperature. Labeled cDNA was cleaned-up using Fairplay columns (Stratagene) according to the instructions of the manufacturer. Labeled reference and fibroblast cDNA samples were combined and mixed with gene expression hybridization buffer and control targets supplied by the manufacturer and hybridized to a 4 × 44K two-color whole human genome gene expression array for 17 hours at 65°C. The array was then washed in a solution of 6× SSPE, 0.005% N-lauroylsarcosine followed by a solution of 0.06× SSPE, 0.005% lauroylsarcosine and scanned on the Agilent DNA Microarray scanner at a resolution of 5 μm. All images were extracted and normalized with Feature Extraction software version 9.5.

The expression arrays for T38 CAF and its counterpart fibroblast were repeated once as technical replicates. The replicated values for each probe in the whole genome expression array were used to calculate a P value using a two-tailed t test comparing each probe's values in T38 CAF to probe values in the counterpart T38 fibroblast. Fold changes between the two samples were calculated using the average probe value in the replicates for each sample.

Microarray data was stored following the MIAME criteria. Microarray results have been submitted to the Gene Expression Omnibus (accession code to be determined shortly).

Surgical tumor specimens from 25 resected invasive breast cancers were minced and fibroblasts were allowed to grow out. We successfully cultured 25 CAFs for at least five serial passages. DNA was extracted from CAFs cultured in passages five to eight. The fibroblast nature of the CAFs was verified by immunohistochemical staining for vimentin and cytokeratin (Supplementary Fig. S1). aCGH was performed on a 244K probe Agilent platform to assess potential changes in DNA copy number. Discrete copy number polymorphisms were detected in many CAFs at regions corresponding to previously annotated copy number polymorphisms (data not shown). These were also detected in normal counterpart fibroblasts. However, only 1 of the 25 CAFs (T38), which was obtained from a triple negative tumor from a 56-year-old patient, showed any non–copy number polymorphism genomic alterations by aCGH and these alterations were present on chromosomes 4, 6, and 9 (Fig. 1). Chromosomes 6 and 9 showed an entire deletion of their respective p arms, whereas chromosome 4 showed a large deletion in the q arm corresponding to loss of cytoband 4q27-30. No amplifications were observed. No alterations, except copy number polymorphisms, were observed in the corresponding counterpart T38 fibroblasts (Supplementary Fig. S2). Karyotyping was also performed on T38 CAF. Due to the cytogenetic complexity of these cells, a total of 38 metaphases were karyotyped and clonal changes are listed in Table 1. The modal chromosome number was between 38 and 46 chromosomes. Copies of several chromosomes (including chromosomes 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, and 22) were missing in at least three metaphases of this CAF (Fig. 2). Several structural rearrangements were also observed.

TP53 mutation screening, initially by single-strand conformation polymorphism of exons 5 to 9, the genomic regions encoding the DNA-binding domain of p53, and then by DNA sequencing of all protein coding exons, to identify mutations missed by single-strand conformation polymorphism analysis, revealed a mutation in only one CAF sample, sample T38. The mutation, a deletion of a guanine residue in exon 4 at nucleotide position 7520207 (relative to GenBank reference sequence NC_000017) results in a frameshift from amino acid position 69 and the introduction of a stop codon at amino acid position 122 (Fig. 3A). This frameshift mutation is predicted to result in the premature termination of the encoded protein that would abrogate the p53 DNA-binding domain. This mutation has been reported at least once in the IARC TP53 Mutation Database version R14 (16). Only the variant allele was detectable by DNA sequencing of T38 (Fig. 3A), and this observation is consistent with the observed LOH of D17S786, a polymorphic microsatellite repeat genetic marker which is located ∼1.2 Mb proximal to the TP53 locus at 17p13.1 (Fig. 3B). LOH of the TP53 locus is also supported by the observation that sequencing of T38 DNA revealed homozygosity for the TP53 IVS3+24 insACCTGGAGGGCTGGGG (ins16) variant, which is heterozygous in patient matched peripheral blood lymphocyte DNA (data not shown). The TP53 mutation was not detected in the analysis of peripheral lymphocyte DNA from the same patient, suggesting that the mutation observed in T38 CAF was the result of a somatic genetic event. We also performed gene expression profiling on Agilent whole genome oligonucleotide expression arrays and found that TP53 (0.75-fold change, P = 2.0 × 10⁻⁶) as well as CDKN1A (0.45-fold change, P = 0.0002), encoding for the p21 protein that is transcriptionally regulated by p53, were among the top 10% of downregulated genes in the T38 CAF compared with the counterpart fibroblast from the same tumor. This indicates a possible functional consequence of the TP53 mutation and LOH.

Interestingly, the T38 CAF was obtained from a patient who had undergone chemotherapy prior to surgical tumor resection, raising the possibility that the chemotherapy could have induced the TP53 gene mutation and/or the associated chromosomal instability. In our cohort, four other CAF samples were also derived from patients who had received preoperative chemotherapy but these CAFs did not show any abnormalities by aCGH or TP53 mutation analysis. Moreover, one of these CAFs (CHUM496) was also karyotyped and its karyotype was found to be normal (Supplementary Fig. S3). Thus, it is unlikely that the chromosomal instability observed in the T38 CAF was related to the administration of preoperative chemotherapy.

Much attention has been focused on the important role of the tumor microenvironment in the development of solid tumors (18). Previous studies have suggested that the presence of genomic alterations similar to those observed in cancer cells (e.g., DNA copy number changes) may facilitate the expression of a tumor-promoting phenotype observed in tumor-associated stromal cells, even when they are removed from tumors and cultured ex vivo. However, our data reinforces the model supported by Qiu and colleagues (10) in which the procancerous effects of CAFs are not due to DNA level alterations in the CAFs themselves, thereby further weakening the notion that such changes are responsible for the stable CAF phenotype. Of the seven cultured breast CAFs profiled in their study, only one showed copy number alterations, occurring on chromosomes 7 and 10. These findings, when considered with those of this present study show that it is possible for DNA copy number–type genomic changes to occur in CAFs, although it does not constitute a common mechanism capable of explaining the CAF phenotype. These results seem to be substantially at odds with mouse models, which showed that CAFs underwent considerable genetic changes in response to tumor formation. For example, Hill and colleagues (5) used a transgenic mouse model of prostate cancer in which a 121–amino acid NH₂-terminal fragment of the SV40 large T-antigen caused mice to develop extensive prostatic intraepithelial neoplasia as a result of the inactivation of pRb and its related proteins. By 72 weeks of age, 7 of 11 mice had developed a Tp53+/Tp53− heterozygous genotype in the stromal compartment, whereas 2 had lost both copies of the wild-type allele. Given the disparity between mouse data and those derived from primary breast cancer samples, the former should be interpreted with caution. There are clearly different factors governing the genetic and genomic fates of the mouse tumor microenvironment, in both the transgenic and xenograft contexts, and that of human cancer patients.

Several reports have also shown the presence of mutations in known tumor suppressor genes such as TP53 and PTEN in the cancer-associated stroma of clinical samples (19). Moreover, the presence of BRCA1/2 mutations in hereditary breast cancers has supported the notion that stromal mutations might act as potential landscaping or facilitating events in the development of subsequent lesions in the adjacent epithelium (20). Additionally, TP53 mutations and LOH events in the cancer-associated stroma of sporadic breast cancers were associated with regional lymph node metastases (9). On the other hand, Qiu and colleagues (10) failed to detect the presence of TP53 mutations by sequencing of exons 4 through 9, the genomic regions encoding the DNA-binding domain of p53, in microdissected areas of stroma adjacent to cancerous epithelia in 10 flash-frozen specimens and 7 cultured CAFs. It is very possible that the majority of stromal mutations detected in human samples in previous studies arose as a result of paraffin-induced DNA artifacts (21). We found only one CAF with a TP53 mutation. Both TP53 and CDKN1A gene expression were underexpressed in this CAF as compared with expression levels in the counterpart fibroblast from the same patient, indicating that the TP53 mutation could be of functional consequence to the CAF. Indeed, this is the first report of a TP53 mutation in a cultured CAF from a human breast tumor. Moreover, this mutation is associated with gross chromosomal abnormalities in the karyotype of the CAF. It is tempting to speculate that the TP53 inactivation enabled the propagation of the unstable karyotype of this CAF, although further work is required to validate this hypothesis.

One of the caveats in studying cells cultured from tumor explants is the possibility that the process of culturing CAFs induced genetic alterations or led to the selection of minor CAF subpopulations. As we analyzed CAFs that were cultured in only five to eight population doublings (or less than four passages), it is unlikely that the TP53 mutation observed in CAF sample T38 arose in cell culture. However, we cannot exclude the possibility that TP53 mutation exerted a selective advantage for cell growth in culturing CAF mutation–positive cells. The consequences of TP53 mutation such as that observed in sample T38 involving several chromosomal abnormalities, is consistent with a mutation background involving this tumor suppressor gene. Thus, whereas cell culturing may have selected a subpopulation harboring these chromosomal abnormalities, the underlying genetic event (TP53 mutation) playing a role in their development was likely present in the initial population of CAFs.

On the other hand, it is possible that a subpopulation of TP53 mutation-positive CAFs and/or CAFs with chromosomal aberrations were masked by genomically normal CAFs, leading to the low detection rate of such CAFs in our series. At present, the sensitivity of mutation screening prohibits the detection of mutations at the single cell level without manipulation of DNA that could introduce genetic anomalies in processing templates for DNA sequencing, whereas the 244K Agilent aCGH platform might be limited to the detection of chromosomal aberrations present in at least 30% of sample cells (22). Although the possibility of minor subpopulations of genomically unstable cells cannot be excluded, it may be expected that cells with a greater propensity to form chromosomal aberrations are more likely to have been selected for in culture conditions, as evidenced by the many reports finding greater numbers of such aberrations in cultured cancer cells than in clinical tumor samples (23, 24).

Given our findings, along with those of others working with fresh and cultured tissue specimens, it is our opinion that breast CAFs should be largely regarded as genomically (i.e., DNA copy number) stable cellular constituents in the complex cancer microenvironment. Importantly, as noted in a recent review (25), the conceptual framework for postulating that CAFs acquire widespread DNA level changes is difficult to accept given that the outcome should preferentially result in a sarcoma or carcinosarcoma type lesion. This is clearly not the case as the incidence of breast carcinomas far outweighs that of mesenchymally derived breast neoplasias. As such, it is much more conceivable that fibroblast to CAF transdifferentiations are due to changes in epigenetic patterning (26) and/or miRNA expression changes in these stromal cells as they are converted into a more reactive phenotype in response to the evolving tumor microenvironment.

No potential conflicts of interest were disclosed.

The authors acknowledge the assistance of Urszula Krzemien, Marie-Claude Huneau, and Marguerite Buchanan in obtaining the human samples.

Grant Support: Sir Mortimer B. Davis Jewish General Hospital Foundation-Weekend to End Breast Cancer Fund and the Fonds de la Recherche en Santé du Québec Réseau de Recherche en Cancer-axe cancer du sein et ovaire.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.